Spaces:
Running
Running
File size: 3,967 Bytes
d70361b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 | """
Resume Day 2-3 tasks SEQUENTIALLY (one heavy job at a time).
Previous crashes were likely caused by running 4 AdaptFormer/TensorFlow jobs in
parallel while also writing thousands of mask PNGs to disk.
Usage:
python scripts/resume_day3_tasks.py
python scripts/resume_day3_tasks.py --from grid
"""
from __future__ import annotations
import argparse
import json
import subprocess
import sys
from pathlib import Path
ROOT = Path(__file__).resolve().parent.parent
PY = sys.executable
def run_step(name: str, cmd: list[str]) -> None:
print(f"\n{'=' * 60}\nSTEP: {name}\n{'=' * 60}")
subprocess.run(cmd, check=True, cwd=ROOT)
print(f"STEP DONE: {name}")
def main():
parser = argparse.ArgumentParser(description="Resume Day 3 tasks sequentially")
parser.add_argument("--from", dest="from_step",
choices=["baseline", "grid", "calibration", "finetune"],
default="baseline")
parser.add_argument("--finetune-epochs", type=int, default=12)
parser.add_argument("--force", action="store_true",
help="re-run steps even if output artifacts already exist")
args = parser.parse_args()
baseline_out = ROOT / "runs/delhi_baseline/metrics.json"
grid_out = ROOT / "runs/calibration/best_params.json"
calib_out = ROOT / "runs/calibration/leaderboard.json"
finetune_glob = ROOT / "runs/finetune_adaptformer"
steps: list[tuple[str, list[str], Path | None]] = [
("baseline", [PY, "scripts/record_delhi_baseline.py"], baseline_out),
("grid", [
PY, "scripts/grid_search_calibration.py",
"--manifest", "docs/delhi_eval/manifest.json",
"--methods", "Feature-Based",
"--sensitivities", "0.2,0.3,0.4,0.5,0.6,0.7,0.8",
"--fusions", "smart_union,hysteresis",
"--out", "runs/calibration/leaderboard.csv",
], grid_out),
("calibration", [
PY, "scripts/delhi_calibration_sweep.py",
"--manifest", "docs/delhi_eval/manifest.json",
"--out", "runs/calibration",
"--methods", "Feature-Based",
"--quick",
], calib_out),
("finetune", [
PY, "scripts/finetune_adaptformer.py",
"--manifest", "docs/delhi_eval/manifest.json",
"--epochs", str(args.finetune_epochs),
"--batch-size", "2",
], None),
]
start = False
for name, cmd, artifact in steps:
if name == args.from_step:
start = True
if not start:
continue
if artifact and artifact.is_file() and not args.force:
print(f"\nSKIP {name}: {artifact} already exists (use --force to re-run)")
continue
if name == "finetune" and not args.force:
existing = sorted(finetune_glob.glob("*/metrics.json"))
if existing:
print(f"\nSKIP finetune: {existing[-1]} already exists (use --force to re-run)")
continue
run_step(name, cmd)
summary = {}
for path, key in [
(baseline_out, "baseline"),
(grid_out, "calibration_best"),
(calib_out, "calibration_leaderboard"),
(ROOT / "runs/calibration/grid_search/manifest_report.json", "grid_search_manifest"),
]:
if path.is_file():
summary[key] = json.loads(path.read_text(encoding="utf-8"))
finetune_runs = sorted(finetune_glob.glob("*/metrics.json"))
if finetune_runs:
summary["finetune"] = json.loads(finetune_runs[-1].read_text(encoding="utf-8"))
out = ROOT / "runs/day3_completion_summary.json"
out.parent.mkdir(parents=True, exist_ok=True)
out.write_text(json.dumps(summary, indent=2), encoding="utf-8")
print(f"\nAll steps finished. Summary: {out}")
if __name__ == "__main__":
main()
|